hog_list: Detect pedestrians using HOGDescriptor

In georoen/wuepix: Using Computer Vision to Count Pedestrians

Description

Detect objects using HOG+SVM (implemented in OpenCV) in all Files/Images of 'path'

Usage

hog(img.folder)

Arguments

winStride	Window stride. It must be a multiple of block stride.
padding	Not implemented yet!
Mscale	Numeric. Allows multi-scale detection. Coefficient of the detection window increase.
resize	Numeric factor resizing image in integrated pre-processing step. E.g. 2 will double the image extent. People should be 100 pixels high.
predictions	dir path to where to store prediction images. Must end with "/".
img.folder	Path to (preprocessed) image archive

Details

Python and OpenCV have to be installed. Tested on Linux only.

Further ideas: [A] Add more 'hog.detectMultiScale' parameters: winStride=(4, 4), padding=(8, 8), scale=1.05) [B] Save predictions.png to a folder

Value

Numeric vector with number of detected persons.

Author(s)

Jeroen Staab

georoen/wuepix documentation built on May 25, 2019, 5:25 p.m.